Data Center Artificial Intelligence (AI) in IndiachipStartups Ceremorphic A life sciences division was opened with the aim of finding and launching drugs based on a new technology platform, BioCompDiscoverX. The platform includes a heterogeneous gas pedal based partly on analog silicon that accelerates the simulation of human cells and tissues as part of the drug development process. The company's goal is to find better candidate molecules from the start, thereby bringing new drugs to market faster and cheaper.
Source Note: The image is generated by AI, and the image is authorized by Midjourney
Ceremorphic moved out of stealth mode in 2022 with the launch of its QS1 chip for data center AI acceleration. The company was founded in 2020 by Venkat Mattela, whose development work began with his previous startup, Redpine Signals, which was acquired by Silicon Labs in 2020.Mattela retained Redpine's AI processor development team and some key intellectual property. The patents related to ThreadArch (Redpine's multi-threaded processor macroarchitecture) have been licensed to Silicon Labs but remain available to Ceremorphic.
Mattela told EE Times that the intention was to pursue drug development from the start. "The problem in life sciences today is that the total number of known diseases is about 10,000, but there are only about 500 available cures." Mattela said, "There is a huge gap between progress or development efficiency and demand."
According to Mattela, the field of life sciences follows the inverse of Moore's Law (known as Eroom's Law); as technology evolves, drug development becomes slower and more expensive. Developing a new drug can now cost more than $2 billion and take more than 10 years, including FDA approval. Last year, the FDA approved only 37 new drugs.
Mattela recalls reading an MIT article about the return of analog computing in drug development in 2016, just before Redpine was sold. This MIT piece of work was about trying to use analog hardware to simulate human cells, or even tissues or organs.
Mattela says: "Using theFirstAdvances in digital computers are not possible. If cells could be simulated more accurately, there would be no need for iterative trials in drug development. When we send 10 drugs into clinical trials, only one makes it through."
Typically, the drug development process may start with identifying a target molecule and optimizing it (target and lead optimization), followed by preclinical testing (animal trials), and finally clinical trials (human trials). mattela says this process now takes more than a decade due to the need for iterative testing. the process can take more than a decade, but it is not as fast as it should be. The current drug failure rate in clinical trials is about 92%. by better mimicking human cells and tissues, each target molecule that passes through the initial development stage has a higher chance of success. the goal of Ceremorphic is not to speed up the clinical trial process itself, but to identify better candidate molecules from the start, so that fewer molecules make it to the trial stage, where expensive and time-consuming trials are more likely to be successful. likely to succeed.
Ceremorphic's key intellectual property lies in the algorithms that simulate human cells, and the hardware gas pedals dedicated to these algorithms. The simulation process addresses the shortage of relevant data, and it will create synthetic data that can be used to train drug discovery AI models with large data requirements, a process Ceremorphic calls Relevant Data Synthesis (RDS).
Mattela's vision for Ceremorphic's BioCompDiscoverX platform is to build a "data center in a box" on every scientist's desk that includes Ceremorphic's silicon chips and can connect to remote resources as needed. Ceremorphic's heterogeneous hardware gas pedals include gas pedals for molecular dynamics, artificial intelligence, and reaction processors, which, crucially, use analog silicon technology.Ceremorphic will remain a fabless semiconductor company, but its new life sciences division will also take on the rest of the value chain, working with partners on clinical trials and manufacturing, although the the exact business model is still evolving.
"Everything is open at the moment", Mattela said. But licensing discovered drugs to pharmaceutical companies could also be part of their business model. He added that similar divisions may emerge in the future for other applications, but that the value proposition is currently insufficient to support such an initiative.
The end result should be the discovery of more drugs and a reduction in the cost of bringing them to market, resulting in a much lower cost to the consumer.Mattela says: "From a social impact perspective, I'm going to be able to get a drug out at the right time so that there's no gap between demand and need. When I increase the probability of success, my costs go down and my time is reduced... Our goal is to personalize drugs as quickly as possible."
Ceremorphic, which currently employs about 160 people, mostly based in Hyderabad, India, plans to demonstrate its technology platform next month, with a full launch scheduled for the end of 2024.